API reference

The AgentRuntime public API is served by the BFF at https://api.agentruntime.io. All customer integrations should use this base URL. For request/response samples, see API examples.

Interactive explorer

Browse a subset of workflow and run endpoints in the marketing site’s Scalar UI: agentruntime.io/api-reference

The marketing OpenAPI spec covers core workflow surfaces. It may lag behind the full BFF route catalog. Cross-check with the tables below or test against api.agentruntime.io.

OpenAPI specifications

Formal OpenAPI 3.1 specs exist for a subset of the API surface, including billing (/v1/billing/*) and onboarding (/v1/onboarding/*, /v1/setup/*) endpoints. There is no single full OpenAPI file covering every route today — use this page, API examples, and the interactive explorer above for the rest of the surface.

Authentication

All authenticated requests use either:

Session cookie — Browser sessions from the Console
Bearer PAT — Authorization: Bearer pat_... for programmatic access

Scoped requests also require:

X-Tenant-Id: your-workspace-slug
X-Project-Id: your-project-id

See API authentication. Role shorthand in tables: viewer+ = project_viewer or higher, contributor+ = project_contributor or higher, tenant_admin = workspace billing admin.

Workflows

Base path: /v1/workflows

Method	Path	Role	Description
`GET`	`/v1/workflows`	viewer+	List workflows
`POST`	`/v1/workflows`	contributor+	Create workflow
`GET`	`/v1/workflows/{id}`	viewer+	Get workflow
`PATCH`	`/v1/workflows/{id}`	contributor+	Update draft graph or run_setup (Save defaults)
`DELETE`	`/v1/workflows/{id}`	contributor+	Archive workflow
`POST`	`/v1/workflows/{id}/versions`	contributor+	Publish version
`GET`	`/v1/workflows/{id}/versions`	viewer+	List versions
`POST`	`/v1/workflows/{id}/validate`	contributor+	Dry-run graph validation
`POST`	`/v1/workflows/{id}/command`	contributor+	Run control (see below)
`GET`	`/v1/workflows/{id}/runs`	viewer+	List runs for workflow
`GET`	`/v1/workflows/{id}/stream`	viewer+	WebSocket event stream

Start, pause, resume, stop

POST /v1/workflows/{id}/command

{
  "command": "start",
  "params": {
    "trigger_type": "api",
    "trigger_payload": {
      "topic": "weekly digest",
      "message": "Hello"
    },
    "trigger_steps": {
      "step-uuid-optional": { "prompt": "per-step override" }
    },
    "use_run_setup_defaults": false
  }
}

Param	Description
`trigger_payload`	Global shared input → `{{input.*}}`
`trigger_steps`	Per-step map (optional overrides / gated entry grants)
`use_run_setup_defaults`	When `true`, merge saved Studio defaults from the workflow version; request keys win on conflicts. Default `false`.

params become workflow input ({{input.*}} in steps). See Run setup.

`command`	`params`	Effect
`start`	Trigger fields	Start a new run
`pause`	`run_id`	Pause scheduling
`resume`	`run_id`	Resume paused run
`stop`	`run_id`, optional `mode`	Stop run (`immediate` default, or `graceful`)

Runs and human tasks

Method	Path	Role	Description
`GET`	`/v1/runs`	viewer+	Tenant-wide run feed (Command Center data source)
`GET`	`/v1/workflows/{id}/runs/{runID}/events`	viewer+	Step event log
`GET`	`/v1/workflows/{id}/runs/{runID}/context`	viewer+	Run variables and step results
`GET`	`/v1/tasks/pending`	viewer+	Pending human tasks (tenant inbox)
`GET`	`/v1/workflows/{id}/runs/{runID}/tasks/pending`	contributor+	Pending tasks for one run
`POST`	`/v1/workflows/{id}/runs/{runID}/tasks/{task_id}/complete`	contributor+	Approve or reject human task

Complete task body:

{
  "approved": true,
  "result": {
    "comment_body": "Optional fields merged into step result"
  }
}

The human task step result always includes approved. The Console Approve/Reject buttons send only { "approved": true|false }. See Human tasks.

Run event types

Persisted on GET /v1/workflows/{id}/runs/{runID}/events. See also Runs and Command Center — Failure handling.

`event_type`	Description
`step_started`	Step began
`step_completed`	Step succeeded, or synthetic skip (`data.skipped: true`)
`step_failed`	Step error; run may transition to `failed`
`step_cancelled`	Step was in flight when another step failed the run
`state_changed`	Run status transition
`run_started` / `run_rejected`	Run lifecycle

Skip reasons on step_completed when skipped: true:

`reason`	Meaning
`blocked`	Run failed before this step started
`out_of_scope`	Episode mode — step outside scope
`step_disabled`	Step disabled in graph

Cancelled step (step_cancelled): data.cancelled: true, data.reason: "run_failed", optional caused_by_step_id. See Runs and Command Center — Failure handling for Console badges.

Inbound webhooks

Method	Path	Auth	Description
`POST`	`/v1/inbound-webhooks`	contributor+	Create subscription
`GET`	`/v1/inbound-webhooks`	contributor+	List subscriptions
`DELETE`	`/v1/inbound-webhooks/{id}`	contributor+	Delete subscription
`POST`	`/v1/inbound-webhooks/{subscription_id}`	HMAC + Bearer PAT	Public ingress — starts run

Create subscription:

{
  "workflow_id": "workflow-uuid",
  "name": "form-leads",
  "automation_pat_id": "optional-stored-pat-id",
  "automation_bearer": "pat_…"
}

Response includes ingress_url, signing_secret (once), and subscription. Ingress request headers:

Authorization: Bearer pat_…
X-Agentruntime-Signature: sha256=<hmac-sha256 of raw body>
Content-Type: application/json

See Inbound webhooks.

Connections

Base path: /v1/connections — project_contributor

Method	Path	Description
`GET`	`/v1/connections`	List workspace connections
`POST`	`/v1/connections`	Create custom connection
`PATCH`	`/v1/connections/{id}`	Update connection
`DELETE`	`/v1/connections/{id}`	Delete connection
`GET`	`/v1/connections/{id}/credentials`	Reveal credential metadata (not full secrets)

Google OAuth

Method	Path	Description
`GET`	`/v1/connections/google/config`	Whether Google connect is enabled
`POST`	`/v1/connections/google/oauth-url`	Start OAuth
`POST`	`/v1/connections/google/complete`	Complete OAuth callback
`GET`	`/v1/connections/google/accounts/{connectionId}`	Account + enabled services
`PATCH`	`/v1/connections/google/accounts/{connectionId}/services`	Enable/disable services

Method	Path	Description
`GET`	`/v1/connections/whatsapp/config`	Whether WhatsApp connect is enabled
`POST`	`/v1/connections/whatsapp/complete`	Complete embedded signup
`POST`	`/v1/connections/whatsapp/manual`	Manual WABA credentials

MCP

Base path: /v1/mcp — project_contributor unless noted.

Servers and catalog

Method	Path	Description
`GET`	`/v1/mcp/platform-catalog`	Platform adapter catalog
`GET`	`/v1/mcp/tools`	Tool catalog
`GET`	`/v1/mcp/servers`	Tenant MCP servers
`POST`	`/v1/mcp/servers`	Register tenant server
`GET`	`/v1/mcp/servers/{id}`	Server detail
`POST`	`/v1/mcp/servers/{id}/validate`	Run validation checks
`POST`	`/v1/mcp/servers/{id}/connections/{connection_id}/validate`	Validate connection binding

Instances and config profiles

Method	Path	Description
`GET`	`/v1/mcp/instances`	List instances
`POST`	`/v1/mcp/instances`	Create instance
`POST`	`/v1/mcp/instances/resolve`	Resolve instances for workflow
`GET`	`/v1/mcp/instances/{id}/config-profiles`	List profiles
`POST`	`/v1/mcp/instances/{id}/config-profiles`	Create profile
`PATCH`	`/v1/mcp/instances/{id}/config-profiles/{profile_id}`	Update profile wiring
`POST`	`/v1/mcp/instances/{id}/config-profiles/{profile_id}/set-active`	Set active profile

Duplicate connection CRUD also exists under /v1/mcp/connections for legacy clients.

AI, chat, and memory

Method	Path	Role	Description
`POST`	`/v1/autopilot/chat`	viewer+	Autopilot message
`GET`	`/v1/autopilot/stream`	viewer+	Streaming response
`GET/POST`	`/v1/chat/*`	viewer+	Chat service proxy
`GET`	`/v1/llm-provider-keys/catalog`	viewer+	LLM model catalog
`POST`	`/v1/llm/resolve`	contributor+	Resolve model ref

Memory (preview)

Memory routes require memory kernel configuration. Console Memory UI may be off in production — see Feature availability.

Method	Path	Description
`POST`	`/v1/memory/search`	Semantic search
`POST`	`/v1/memory/index-conversation`	Index chat episode
`POST`	`/v1/memory/index-file`	Index work file
`POST`	`/v1/memory/extract/batch`	Batch extract jobs
`GET`	`/v1/memory/extract/{jobId}`	Extract job status
`POST`	`/v1/memory/tools/call`	Agent memory tool call

Billing endpoints

Base path: /v1/billing

Method	Path	Role	Description
`GET`	`/v1/billing/summary`	contributor+	Tenant/project summary
`GET`	`/v1/billing/plans`	contributor+	Plan catalog for this deployment’s `product_code` (Stripe-synced)
`GET`	`/v1/billing/status`	contributor+	Subscription and trial status
`GET`	`/v1/billing/usage`	contributor+	Credit balances (microcredits)
`GET`	`/v1/billing/usage-dimensions`	viewer+	Usage breakdown dimensions
`GET`	`/v1/billing/usage-history`	contributor+	Metered usage ledger
`GET`	`/v1/billing/payment-history`	contributor+	Payments and top-ups
`GET`	`/v1/billing/action-pricing`	contributor+	Resolved per-action prices
`POST`	`/v1/billing/topups/payg/intent`	contributor+	PAYG top-up Stripe intent
`POST`	`/v1/billing/trial`	tenant_admin	Start trial
`POST`	`/v1/billing/subscription`	tenant_admin	Activate subscription
`POST`	`/v1/billing/subscription/change-plan`	tenant_admin	Change plan
`POST`	`/v1/billing/subscription/cancel`	tenant_admin	Cancel at period end (default)
`POST`	`/v1/billing/subscription/convert-trial-now`	tenant_admin	End trial early
`POST`	`/v1/billing/codes/apply`	tenant_admin	Apply billing code
`POST`	`/v1/billing/discounts/validate`	tenant_admin	Validate discount

GET /v1/config/stripe — publishable Stripe config for Console checkout. GET /v1/billing/plans returns tiers scoped to your workspace’s product (AgentRuntime Console → agentruntime). Catalog rows are keyed by (product_code, plan_code); the public response surfaces plan_code and pricing fields. See Multi-product billing. See Billing and credits and API examples.

Agents

Base path: /v1/agents — project_viewer (read), project_contributor (write). Requires enableAgents in your deployment.

Method	Path	Description
`GET`	`/v1/agents`	List agents
`POST`	`/v1/agents`	Create agent
`GET`	`/v1/agents/{id}`	Get agent
`PATCH`	`/v1/agents/{id}`	Update agent
`DELETE`	`/v1/agents/{id}`	Delete agent

Analytics

Base path: /v1/analytics — project_contributor

Endpoint	Metric
`/v1/analytics/live-runs`	Currently running workflows
`/v1/analytics/success-rate`	Run success percentage
`/v1/analytics/execution-duration`	Wall-clock execution time
`/v1/analytics/runs-count`	Runs in selected window
`/v1/analytics/steps-executed`	Metered workflow steps
`/v1/analytics/credits-consumed`	Credits used in window
`/v1/analytics/steps-summary`	Step-type breakdown and latency
`/v1/analytics/runs-timeseries`	Run count over time
`/v1/analytics/usage-timeseries`	Usage over time

Query window=month|week|all or bounded from / to with optional granularity on timeseries endpoints. See Analytics.

Personal access tokens

Method	Path	Role	Description
`GET`	`/v1/pats`	contributor+	List tokens
`POST`	`/v1/pats`	contributor+	Create token
`GET`	`/v1/pats/{id}/reveal`	viewer+	Reveal token (once)
`DELETE`	`/v1/pats/{id}`	contributor+	Revoke token

Console Settings → API keys manages the same keys for inbound webhook automation.

Error codes

Status	Meaning
`401`	Missing or invalid authentication
`403`	Insufficient project role or PAT scope
`404`	Resource not found
`422`	Validation error (graph, bindings, billing)
`429`	Rate limited
`5xx`	Server error — retry with backoff

SDKs and collections

Request/response samples — see API examples for copy-paste requests
Platform MCP — mcp.agentruntime.io; see Platform MCP

Contact support@agentruntime.io for enterprise API SLAs.

Get started

Platform

Workflows

Guides

Integration setup

Connectors

AI & chat

API

Interactive explorer

OpenAPI specifications

Authentication

Workflows

Start, pause, resume, stop

Runs and human tasks

Run event types

Inbound webhooks

Connections

Google OAuth

WhatsApp

MCP

Servers and catalog

Instances and config profiles

AI, chat, and memory

Memory (preview)

Billing endpoints

Agents

Analytics

Personal access tokens

Error codes

SDKs and collections

​Interactive explorer

​OpenAPI specifications

​Authentication

​Workflows

​Start, pause, resume, stop

​Runs and human tasks

​Run event types

​Inbound webhooks

​Connections

​Google OAuth

​WhatsApp

​MCP

​Servers and catalog

​Instances and config profiles

​AI, chat, and memory

​Memory (preview)

​Billing endpoints

​Agents

​Analytics

​Personal access tokens

​Error codes

​SDKs and collections

Interactive explorer

OpenAPI specifications

Authentication

Workflows

Start, pause, resume, stop

Runs and human tasks

Run event types

Inbound webhooks

Connections

Google OAuth

WhatsApp

MCP

Servers and catalog

Instances and config profiles

AI, chat, and memory

Memory (preview)

Billing endpoints

Agents

Analytics

Personal access tokens

Error codes

SDKs and collections